OpenAI kills Sora as compute costs force a strategic retreat

OpenAI announced on Tuesday that it is discontinuing Sora, its AI video generation platform, just six months after the standalone app launched. The shutdown covers the consumer app, the developer API and video functionality within ChatGPT. CEO Sam Altman informed staff that the company would wind down products built on its video models.^[1]

OpenAI has since published the shutdown timeline: the Sora web and app experiences are set to end on April 26, 2026, and the Sora API on September 24, 2026. The company says users should export their content before the web and app shutdown date.^[2]

The decision also ends OpenAI's partnership with Walt Disney Co., announced in December 2025, which would have licensed more than 200 Disney characters for use in Sora and included a $1 billion Disney investment in OpenAI.^[1][3] The deal was reportedly structured entirely in stock warrants rather than a cash licensing fee^[4], and no money ever changed hands.^[5]

The announcement blindsided Disney. On Monday evening, Disney and OpenAI teams were still collaborating on a Sora-related project. Thirty minutes after that meeting, the Disney team learned OpenAI was dropping the tool entirely. A person familiar with the matter described it as "a big rug-pull".^[6]

A product that couldn't sustain its own launch

Sora 2 and its standalone iOS app launched on September 30, 2025.^[7] The app reached one million downloads within five days and rose to the top of the iPhone App Store. An Android version followed later that year.^[9]

But the initial surge did not hold. Data from mobile intelligence firm Appfigures shows that Sora's downloads peaked at approximately 3.3 million in November 2025 [11], then dropped 32% month-over-month in December, a period when app downloads typically rise due to holiday device sales. The decline accelerated in January 2026, with installs falling 45% month-over-month to 1.2 million.^[10] By February, monthly downloads had declined to approximately 1.13 million.^[11]

Consumer spending followed the same trajectory. January revenue from in-app purchases dropped to $367,000, down from a December peak of $540,000.^[12] Total lifetime revenue from in-app purchases was approximately $2.1 million.^[11]

That is $2.1 million from a product inside a company valued at $730 billion, consuming computational resources that OpenAI's own head of Sora, Bill Peebles, had already acknowledged were constrained enough to require generation limits.^[13]

The compute problem

Video generation is far more resource-intensive than text or code generation. Running the Sora app required significant computational resources that left other teams with less firepower.^[6] Some OpenAI employees were surprised by the amount of computing resources the company poured into the project, given the lack of clear evidence of demand.^[1]

OpenAI's spokesperson framed the pivot in terms of research priorities, saying the Sora research team would continue to focus on "world simulation research to advance robotics that will help people solve real-world, physical tasks".^[14] But the broader message from leadership was more direct. In an all-hands meeting earlier this month, Fidji Simo, OpenAI's chief of applications, told employees they could not afford to be distracted by "side quests" and that the company is "orienting aggressively" towards high-productivity use cases.^[8]

A wider consolidation

Sora was not the only casualty on Tuesday. OpenAI also announced it would end its Instant Checkout shopping feature.^[8] Last week, the company said it would combine its ChatGPT desktop app, Codex coding tool and Atlas web browser into a single application.^[1] The Sora shutdown is part of a broader effort to redirect OpenAI's computing resources and top talent toward productivity tools.^[1]

The timing is not incidental. OpenAI recently completed a $110 billion funding round at a valuation of approximately $730 billion.^[13] An initial public offering is widely expected as soon as the fourth quarter of this year.^[1] Shutting down a compute-intensive product with declining engagement and minimal revenue is a straightforward way to clean up the books ahead of opening them to public market investors.

The competitive pressure

The Sora shutdown has been widely framed in the context of competition from Anthropic. Anthropic has eschewed products like image and video generation, instead focusing scarce computational resources on text and code generation. The enterprise productivity space where OpenAI is now pivoting is specifically the area where Anthropic has built a significant business.

This is where the story becomes more than a single product shutdown. OpenAI is converging on the same resource allocation posture that Anthropic has maintained from the outset: concentrate scarce GPU capacity on text, code and reasoning rather than spreading it across experimental consumer products. Every chip rendering a Sora video was one not available for training or serving the reasoning and coding models that enterprise customers are actually paying for.

The strategic logic is straightforward. Sora put OpenAI in a two-front competition it could not win: fighting Google on video generation, where Google has deeper pockets and owns the world’s biggest video distribution channel (YouTube), while simultaneously losing ground to Anthropic on the productivity and enterprise tools that generate sustainable revenue. Something had to give.

What comes next

Altman told staff the Sora team would now give priority to longer-term bets such as robotics. OpenAI has also recently completed development of a new AI model codenamed Spud, which Altman told staff "can really accelerate the economy". The model's capabilities have not been publicly detailed.

For Disney, the collapse of the partnership leaves an open question about which AI platform, if any, major IP holders will next bet on for generative video. Disney's statement was diplomatic, saying it "will continue to engage with AI platforms to find new ways to meet fans where they are while responsibly embracing new technologies that respect IP and the rights of creators.”^[3]

For the broader AI industry, Sora's trajectory from record-breaking launch to shutdown in six months is a concrete illustration of a principle that is becoming harder to ignore: in a compute-constrained environment, focus is a competitive advantage. The company that started the generative AI race by trying to do everything is now retreating to the same strategic posture as its most disciplined competitor.